1. Field of the Invention
The present invention relates to non-volatile semiconductor memory devices, and more particularly, to a MOS-type semiconductor memory device which allows electrical writing and erasing, and a fabrication method therefor.
2. Description of the Related Air
FIG. 15 shows a cross-sectional structure of a memory cell in a conventional NOR-type flash memory. In FIG. 15, there are shown a p-type silicon semiconductor substrate 1, a field oxide film 2 having a thickness of about 400 nm to 800 nm, an oxide film 3 about 10 nm thick formed by thermally oxidizing the silicon semiconductor substrate, serving as a first gate insulating film of the non-volatile memory, a floating gate 4 made from a first polycrystalline silicon film, a second gate insulating film 5 about 20 nm thick, a control gate 6 made from a second polycrystalline silicon film, an impurity diffusion layer 7 having a conductivity type opposite to that of the substrate used for making a source diffusion layer have a high breakdown voltage, an n.sup.+ diffusion layer for the source 8, an n.sup.+ diffusion layer for the drain 9, a side-wall insulating film 10 for a gate electrode, an interlayer insulating film 11, a metal-wiring lead-out electrode 12 for the source, a metal-wiring lead-out electrode 13 for the drain, and a metal-wiring lead-out electrode 14 for the control gate. In a writing operation of the non-volatile semiconductor memory device having such a structure, a voltage is applied between the control-gate electrode 14 and the drain electrode 13 to generate an avalanche hot electron between the drain 9 and the substrate 1, and the electron is injected into the floating gate 4. In an erasing operation, a voltage is applied between the control-gate electrode 14 and the source electrode 12, and the electron is drawn from the floating gate 4 to the source 8 by Fowler-Nordheim tunneling (or F-N tunneling) current.
As scaling-down of an MOS transistor is further progressed to increase an integration scale of the MOS memory devices, the conventional memory cell will have no choice but encounter the following drawbacks:
(1) Since it is said that the lowest limit of a thin-film for a tunneling oxide film is about 8 nm thick in a memory cell using an F-N tunneling current, a scaling rule (a proportional reduction rule between a device dimension and a voltage to be applied, in scaling down) is not satisfied for a thin film having a film thickness less than that. PA1 (2) When an effective channel length is made smaller by scaling-down, it is difficult to control a punch-through breakdown voltage and a leakage current. PA1 (3) As capacitance of a capacitor coupled with the floating gate is reduced, erasing and writing characteristics deteriorate. These three drawbacks will be described below in detail.
FIG. 1 shows a capacitor model in a floating-type semiconductor memory cell. With this model, floating gate voltages Vfg(W) for writing and Vfg(E) for erasing can be expressed, respectively, as follows:
Writing
Vfg(W)=(C.sub.2.multidot.Vcg+C.sub.3.multidot.Vd)/(C.sub.1 +C.sub.2 +C.sub.3 +C.sub.4) (1)
Erasing EQU Vfg(E)=(C.sub.2.multidot.Vcg+C.sub.4.multidot.Vs)/(C.sub.1 +C.sub.2 +C.sub.3 +C.sub.4) (2)
where, C.sub.1, C.sub.2, C.sub.3, and C.sub.4 indicate the capacitance of capacitors formed between the floating gate 4 and the substrate 1, the control gate 6 and the floating gate 4, the floating gate 4 and the diffusion-layer drain 9, and the floating gate 4 and the diffusion-layer source 8, respectively, and Vcg, Vs, and Vd indicate the control gate voltage, the source voltage, and the drain voltage, respectively.
In general, to increase writing efficiency, since Vcg and Vd are both positive values, it is desirable that the capacitance of the capacitors C.sub.2 and C.sub.3 are made as large as possible in expression (1). Then, the effective voltage Vfg(W) applied to the floating gate increases and the injection efficiency of the avalanche hot electron into the floating gate, namely, writing efficiency, is improved. As a result, a writing time becomes short.
On the other hand, during erasing, since Vcg is zero or a negative value and Vs is a positive value, capacitance relationship of C.sub.2 &gt;&gt;C.sub.4 is the most appropriate condition in expression (2). When this condition is satisfied, the effective voltage Vfg(E) applied to the floating gate increases. Therefore, the drawing efficiency of the electron from the floating gate to the source by the use of the F-N tunneling current, namely, erasing efficiency, is improved. As a result, an erasing period becomes short.
From the above reasons, when a device is scaled down finer only according to the scaling rule without changing the shape of MOS transistors, a range of choice in capacitance becomes narrower. Therefore, a non-volatile semiconductor memory device free from the above drawbacks in scaling-down has been demanded.